Media analyst Ian Whittaker argued in a LinkedIn post dated August 28, 2026 that the advertising industry's 2026 growth forecasts carry an unpriced dependency on artificial intelligence data centre construction, a build programme now facing 75% voter opposition in the United States and at least $156bn in blocked or delayed projects.

The argument is narrow and it is specific. Advertising forecasts have grown more optimistic through 2026 on the strength of one input: capital flowing into AI infrastructure, and the advertising that capital generates. Whittaker's contention is that this input is not a technology variable at all. It is a planning-permission variable, and forecasters have historically been poor at pricing planning permission.

"Advertising forecasts may have a data centre problem," the post opens.

An analyst with the forecasting model in hand

Whittaker is founder and managing partner of Liberty Sky Advisors and a former City equity research analyst, twice named City AM Analyst of the Year. His LinkedIn profile describes his work as advising chief executives, chief financial officers, chief marketing officers and investors on the economics and capital allocation shaping media, marketing and technology.

The credential he invokes in the post is more particular than the title. One of his previous roles, he writes, involved redesigning the global advertising forecasts for one of the agency groups, which gives him what he calls a reasonable idea of how much judgement sits underneath the published numbers. That framing matters to how the rest of the argument reads. This is not a critique of forecasting from outside the discipline.

Whittaker has been a recurring source of contrarian readings on advertising economics through 2026. In May he argued that the global advertising market is structurally bifurcated into a visible, agency-mediated layer measured by trade forecasts and a much larger, faster-growing self-serve layer that those forecasts largely miss. In August he set out a separate case that the regulatory template closing around large platforms resembles the conduct regime once imposed on classified directories rather than antitrust divestiture. On a London forecast panel on July 7, 2026, he offered a blunter verdict, telling the audience that advertising has lost its way.

The forecast at the centre of the argument

The post anchors on WPP Media's midyear upgrade, published in June 2026. Whittaker describes the revision as taking the 2026 global forecast to plus 8.9%, or around $1.3 trillion, excluding United States political spend, which he notes means the number is not being flattered by the midterm cycle. He states that WPP Media identified AI investment "and the advertising it generates" as the countervailing force offsetting geopolitical drag.

There is a factual discrepancy here that is worth setting out plainly rather than smoothing over. WPP Media's This Year Next Year midyear forecast, released on June 16, 2026 and authored by Kate Scott-Dawkins, projected 4.4% global advertising revenue growth to $1.3 trillion, with United States growth at 11.9%. The 8.9% figure belongs to a different house. WARC revised its global ad growth forecast upward by 1.2 points to 8.9% for 2026, also on June 16, 2026. The two numbers use different methodological baselines, with WPP Media excluding United States political advertising.

The dollar total and the attribution of causation are consistent across both readings. The growth rate is not. Whittaker's central claim survives the mismatch, because both forecasts credit the same driver, but the specific percentage he cites does not correspond to the WPP Media publication he names.

What is not in dispute is the mechanism WPP Media described. AI-native companies building large language models, inference infrastructure and agent frameworks are buying advertising to reach developers, enterprise buyers and consumers. Traditional advertisers are simultaneously deploying AI to compress production costs and widen activation. WPP Media called the combination a powerful countervailing force to external economic headwinds.

Whittaker does not dispute that reading. "I do not think that is wrong," he writes. His point is what follows from it: a good chunk of the upgrade is, in effect, a bet that the AI buildout carries on. And the buildout, in his phrasing, "is now running into the one thing forecasters have never been much good at pricing, which is politics."

Four polls, one direction

The evidence Whittaker assembles for the political constraint is a polling series rather than a single reading. Heatmap has asked American voters four times in a year whether they would support or oppose a data centre near their home.

A year ago, 42% were opposed and 44% were in favour. Opposition reached 52% in February. It reached 71% in May. It stood at 75% in the month of the post, with more than six in ten of those opposed describing themselves as strongly opposed.

That is a 33-point move in twelve months, and it crossed from net favourable to net hostile somewhere in the winter. Rural voters, a Republican-heavy group, are particularly hostile, according to the post. That detail carries weight because it removes the partisan escape hatch: this is not opposition concentrated in jurisdictions where large infrastructure projects already face long approval timelines.

Intensity matters as much as direction. A margin of six in ten opposed voters holding a strong position is the profile that produces county commission turnouts, zoning appeals and ballot measures rather than unfavourable poll numbers that never translate into procedural friction.

The money already stopped

Polling would be a soft indicator on its own. Whittaker pairs it with a hard one. Data Center Watch counts at least $156bn of projects blocked or delayed by local opposition last year.

That figure is the load-bearing number in the argument. It converts sentiment into schedule risk and capital risk, and it does so retrospectively, covering a period during which opposition was measurably lower than it is now. If $156bn of projects met local resistance while national opposition sat closer to 50%, the base rate at 75% is unlikely to be lower.

The post identifies electricity as the reason. One study in Environmental Research Letters, Whittaker writes, puts wholesale power prices in the worst affected regions up as much as 57% by 2030, against a national average of 6 to 29%.

The contested causal chain

The electricity claim is where the argument meets its most substantial counter-evidence, and the record deserves to be set out rather than assumed.

Two studies published in May 2026 pushed back on the proposition that data centres are directly responsible for rising household electricity costs. PwC, commissioned by the Data Center Coalition, calculated that the industry supported 5.5 million United States jobs and $927bn in GDP in 2024. Energy consultancy E3 examined the rate question directly. Its analysis found no clear correlation between load growth and rate increases at state level. Texas and Virginia, which absorbed the largest increases in electricity load, recorded some of the smallest rate increases across the 2019 to 2024 period. California and New York saw the largest price increases while load actually declined. A Lawrence Berkeley National Laboratory study cited in the same work found nominal electricity prices rose 29% nationally between 2019 and 2025, with PJM as the principal exception to the broader pattern.

Those findings do not contradict Whittaker's argument, because his argument does not require the causal chain to be true. It requires voters to believe it. Rate increases have been real and national; attribution to data centres has been local and vivid. Political opposition responds to the second, not the first.

That distinction is the analytically useful part of the post. A forecaster modelling advertising revenue from AI capital expenditure needs the projects to be built. Whether the projects deserve the blame they attract in county meetings is irrelevant to whether the meetings stop them.

Why an advertising forecast sits downstream of a zoning hearing

The transmission mechanism between data centre construction and advertising revenue runs in both directions, which is what makes the exposure larger than it first appears.

In one direction, AI companies buy advertising. That is the revenue stream WPP Media credited. In the other, advertising revenue funds the construction. Alphabet raised approximately $85 billion in equity in June 2026 to fund infrastructure it described as supply-constrained, combining a $10 billion private placement to Berkshire Hathaway, a $30 billion underwritten public offering and a $40 billion at-the-market programme. On July 22, 2026 the company raised 2026 capital expenditure guidance to between $195 billion and $205 billion, reported negative free cash flow of $5.9 billion for the quarter and disclosed $98.2 billion in long-term debt, against roughly $16 billion a year earlier.

Meta reported quarterly capital expenditures of $31.08 billion on July 30, 2026, against $17.01 billion a year earlier, with free cash flow falling to $784 million while advertising revenue climbed 27% to $59.36 billion. Its full-year guidance sits at $125 billion to $145 billion. Microsoft capital expenditures rose 70% to $41.0 billion in its June quarter. Amazonadvertising grew 26% to $19.8 billion in the same period. S&P Global Ratings cut Oracle to BBB- in July 2026 over a forecast $42bn free operating cash flow deficit, with the agency noting that the unit economics of the AI infrastructure business remain opaque.

These are the same companies that sell most of the world's advertising. A delay to a gigawatt campus is not an abstraction for media planners; it appears in the capital expenditure line of the platforms that set auction dynamics, and eventually in the advertising budgets of the AI-native companies competing for compute they cannot yet secure.

Financing conditions have already shown they can move independently of the published spending forecasts. a debt-financed fund position built on AI infrastructure lost roughly two-thirds of its value in a single month during July 2026 without any change to those forecasts, which located the risk in financing rather than in demand assumptions. Local opposition is a second such channel, operating on physical delivery rather than on capital markets.

The concentration problem

WPP Media projected 11.9% advertising revenue growth in the United States for 2026, well above the 4.4% global average, and attributed the gap to the concentration of AI infrastructure investment in the American market.

That concentration is precisely where the polling risk sits. The Heatmap series measures American voters. Data Center Watch counts American projects. The forecast component most exposed to the buildout continuing is the same component most exposed to the buildout being obstructed.

The timing compounds it. The United States midterm cycle runs through November 2026, with political connected TV spending projected at $2.7 billion and broadcast at $5.6 billion according to AdImpact estimates. Data centre siting has become a live campaign issue in states hosting significant capacity. Neither WPP Media's headline figure nor Whittaker's cited 8.9% includes political advertising, so the midterm spend does not flatter either number. What the cycle does supply is a mechanism for turning the polling into policy.

What forecasters can and cannot model

The methodological point underneath the post is not that agency forecasts are careless. It is that they are built to model demand.

Advertising forecasts extrapolate from category spend, macroeconomic indicators, media inflation, channel migration and platform disclosure. Those inputs have well-understood error bands. Permitting outcomes, utility commission rulings and county-level zoning decisions have none of that structure. They are lumpy, jurisdiction-specific and correlated with each other only through sentiment.

The relevant exposure is therefore not a percentage point of error on a growth rate. It is a category of risk sitting outside the model entirely, weighted at zero by construction rather than by judgement.

Water constraints add another dimension that has surfaced repeatedly in local disputes. PPC Land has documented the effect of Meta's data centre construction on the local water supply in Newton County, and Google's environmental disclosures for 2025 recorded a 37% increase in electricity demand alongside a 25% rise in supply chain emissions and water replenishment reaching 78% of freshwater consumption.

Why this matters for the marketing community

Budget planning for the remainder of 2026 and for 2027 rests substantially on forecast documents that share a common assumption. The assumption is that AI capital expenditure continues on the trajectory the platforms have guided, and that the advertising it generates continues with it.

The practical consequence is a question about forecast dependency rather than about data centres. Media plans built on 8.9% or 11.9% growth assumptions are, at one remove, plans built on construction schedules in Georgia, Texas, Virginia and Ohio. The IAB projected 9.5% United States advertising growth in January 2026, before the polling reached its current level. Meta, Alphabet, Amazon and Microsoft collectively exceeded $150 billion in Q1 2026 advertising revenue, and Alphabet's own Q2 2026 results showed Search and Other advertising rising 17% to $63.3 billion while Network revenue fell 1%.

The infrastructure exposure is not confined to the United States. Amazon committed 33.7 billion euros to data centre expansion in Spain in March 2026, a programme that carries its own permitting and grid-connection dependencies in European jurisdictions where public opinion on the same question has not been polled to comparable depth.

Whittaker's post does not forecast a downgrade. It identifies an assumption that has been carried without being examined, and it puts three numbers against that assumption: 75% opposition, $156bn already blocked or delayed, and a potential 57% regional increase in wholesale power prices by 2030. Whether those numbers eventually appear in a revision is a question for the December forecast round.

Timeline

Summary

Who: Ian Whittaker, founder and managing partner of Liberty Sky Advisors, a former City equity research analyst and twice City AM Analyst of the Year, who previously redesigned the global advertising forecasts for an agency group. The organisations referenced in his analysis include WPP Media, Heatmap, Data Center Watch and the journal Environmental Research Letters.

What: A LinkedIn post arguing that 2026 advertising forecasts contain an unexamined dependency on AI data centre construction continuing, and that this dependency is exposed to local political opposition rather than to technology or demand risk. The supporting figures are a Heatmap polling series showing opposition to nearby data centres rising from 42% to 75% across twelve months, a Data Center Watch count of at least $156bn in projects blocked or delayed by local opposition during 2025, and an Environmental Research Letters finding of wholesale power price increases of up to 57% by 2030 in the worst affected regions against a national range of 6 to 29%.

When: Published on LinkedIn on August 28, 2026, and subsequently edited. The forecast it examines was published in June 2026. The polling series spans August 2025 to August 2026.

Where: The polling and the blocked-project count cover the United States, where AI infrastructure investment is most concentrated and where WPP Media projects 11.9% advertising revenue growth for 2026 against a 4.4% global average. The post itself was published on LinkedIn.

Why: The argument matters to advertising planning because the forecasts underpinning 2026 and 2027 budget assumptions attribute their upgrades to AI investment and the advertising it generates. If data centre construction slows for political reasons, the advertising revenue modelled from that investment slows with it, and that risk sits outside the input set that agency forecasting models are built to capture. A separate discrepancy surfaces in the post itself: the 8.9% growth figure it attributes to WPP Media corresponds to WARC's revised estimate, while WPP Media's own midyear forecast projected 4.4% growth to the same $1.3 trillion total.